That's a fair point and I sure won't be discussing that - I'm still rocking Core 2 Quad at work, because once these oldies are fed a 64-bit Windows, 8 Gb of RAM and a SATA SSD, they really do well at office stuff. And yes, a 3770k is even better at that - or a 4670k (what I used to have).As an employee, would you personally throw $600 at a PC to improve your productivity by 3% by reducing the the 10% of the time you are waiting for your PC for things you can actually just sit in front of the PC and wait for by 30%? No, you'd just keep using whatever the company provided and unless it is grossly inadequate, you wouldn't mind it much. A 30% improvement still does not really matter to you directly, it only matters to the company if it thinks it'll get that much extra productivity out of you. For any operations longer than that, most companies have no shortage of ancillary work to do in the meantime and the net productivity gain is even lower.
The stock Ryzen 1400 would be the VERY POOR man's i7 since the i5-3470 is still up to 20% faster in single-threaded performance when not using AVX2 instructions (not supported by IB) and only ~10% slower in multi-threaded benchmarks despite the 1400's SMT advantage. An actual i7 would be another 20-30% faster on top of that.And don't tell me my best case is a stretch! Because the R5 1400 was pretty much a poor man's i7 when it came out, and as such was a nice pick if you wanted to get a "modern" machine on the cheap back then. Worst case though, going from a 1700X to a 3900X would yield "only" 70% more CPU power - but then the only way to get more oomph would require one to get into HEDP territory.
I don't know where you pulled your 20% extra speed from (UserBenchmark puts it at 10%, 12% tops in single threaded), especially now that all the Spectre/Meltdown patches have been bogging it down; if, to you, an 'actual' i7 is the 7700k from pretty much the same time, then yes, an extra GHz clock speed and 15% better IPC did give the i7 7700k a 40+% performance lead - for 2.7 times the price. Heavy (especially when you consider that the i7 comes with no cooler and needs a pricy motherboard to run).The stock Ryzen 1400 would be the VERY POOR man's i7 since the i5-3470 is still up to 20% faster in single-threaded performance when not using AVX2 instructions (not supported by IB) and only ~10% slower in multi-threaded benchmarks despite the 1400's SMT advantage. An actual i7 would be another 20-30% faster on top of that.
In the case of most user-interactive software like CAD and games, having 200 cores is of limited use since the user interaction management is usually a single-threaded bottleneck on everything else when the user isn't, hence the desktop CPU market's still heavy emphasis on single-threaded performance despite more simpler slower cores (GPU-like) being far more cost and power efficient.
When you need a ton of processing power for embarrassingly parallel stuff, nothing beats GPU acceleration for the price. $2000+ GPUs may sound expensive but they are still much cheaper per TFLOP than desktop CPUs.when you need to render your project, nothing beats a bunch of CPU cores casting rays left and right - and there, nothing in the x86 world of 2017 beat a nice old R7 1700 for the price.
Provided the app you're using supports using a GPU for rendering - which is not always the case, as surprising as it may sound.When you need a ton of processing power for embarrassingly parallel stuff, nothing beats GPU acceleration for the price. $2000+ GPUs may sound expensive but they are still much cheaper per TFLOP than desktop CPUs.
As GPU-accelerated stuff becomes increasingly prevalent, whatever software still doesn't support GPU acceleration will lose more customers to software that does.Provided the app you're using supports using a GPU for rendering - which is not always the case, as surprising as it may sound.
These very specific applications that cost thousands of bucks per seat and have 10 to 20 years of legacy code embedded in them don't usually adopt new stuff fast, and are even slower at dropping stuff - there's a reason why they forced Khronos to keep a Compat mode in OpenGL.As GPU-accelerated stuff becomes increasingly prevalent, whatever software still doesn't support GPU acceleration will lose more customers to software that does.
Lol core per dollar stagnation, we were at 4 core Max for what 10 years?No need to wait for reviews to be disappointed with core-per-dollar stagnation. Assuming the best IPC and clock gains possible, we're still talking only 25-30% more performance per dollar than two years ago, which isn't particularly exciting either. Sure, this is better than Intel's 5-7%/year for most of the past eight years, but not something most people will find worth bothering with if they already have anything somewhat recent.
I thought I'd be upgrading to a 3600 myself, turned out it wasn't the CPU I was hoping for. Maybe I'll get a 3700X if it drops below $200 next year.
Until eight years ago, performance still doubled every 2-3 years and prices for that performance were still dropping. My disappointment is with the brakes coming off after 10 years only to come back on for the next 4+ years.Lol core per dollar stagnation, we were at 4 core Max for what 10 years?
Until eight years ago, performance still doubled every 2-3 years and prices for that performance were still dropping. My disappointment is with the brakes coming off after 10 years only to come back on for the next 4+ years.
While it may be nearly impossible to increase single-thread IPC much further and clock frequencies have hit what appears to be a practical ceiling at ~5Ghz, there is plenty of room left to improve multi-threaded performance. If AMD had wanted to press its current advantage for all it is worth, I have no doubt it could have made the 3600 a $200 8c16t part and still turned more profit than it has for most of the past 10 years while drastically turning up the heat on Intel. Since Intel is going to be effectively MIA in the desktop space for the next 2-3 years though, AMD has every incentive to hold back so it can maximize profit instead of massively increasing market share.I feel like thats more because AMD took a good long while with Zen in the oven in order to not have a repeat of bulldozer, resulting in such a significant increase in performance. Now that they are releasing essentially yearly (or every 2 years) releases of new processor generation, I simply can't expect as much performance increase from Zen to Zen2 or Zen2 to Zen3 as I did from Bulldozer to Zen. It is a shame that they are more or less following the Intel method of tick/tock, but I still think the overall rate is up from the past 8-10 years.
While it may be nearly impossible to increase single-thread IPC much further and clock frequencies have hit what appears to be a practical ceiling at ~5Ghz, there is plenty of room left to improve multi-threaded performance. If AMD had wanted to press its current advantage for all it is worth, I have no doubt it could have made the 3600 a $200 8c16t part and still turned more profit than it has for most of the past 10 years while drastically turning up the heat on Intel. Since Intel is going to be effectively MIA in the desktop space for the next 2-3 years though, AMD has every incentive to hold back so it can maximize profit instead of massively increasing market share.
AMD could have delivered way more than 20-30% more performance per dollar this year and completely obliterate Intel's lineup for the foreseeable future, it simply chose not to.
The chipset itself is only ~$10 more, give it a few months for the dust to settles on those ridiculously overbuilt halo launch products for early adopter enthusiasts, then we should see more reasonable x570 boards that cost only $20 or so more than their x470 counterparts pop up.The 500 series supposedly going to be expensive to get the full potential of the 3000's.
It’s not really the brakes coming on you can expect them to offer 16 cores the gen after 8. They’ve bumped it up to 12 with the 3900xUntil eight years ago, performance still doubled every 2-3 years and prices for that performance were still dropping. My disappointment is with the brakes coming off after 10 years only to come back on for the next 4+ years.
Not particularly helpful for performance per dollar as when you move up the product stack, performance per dollar almost always gets worse. Also, moving up the product stack is only an option if you have no specific budget (price point) in mind.It’s not really the brakes coming on you can expect them to offer 16 cores the gen after 8. They’ve bumped it up to 12 with the 3900x
They’re not going to sell a 12c CPU for 5 quid mate. 8 cores is relatively new, you get a decent bump with the shrink plus higher clocks across the board. You’re getting around intel IPC and 6cores 12 threads for 200 compared to 4 cores 4 threads for 250 in the 7600K or 4/8 for 350 with the 7700KNot particularly helpful for performance per dollar as when you move up the product stack, performance per dollar almost always gets worse. Also, moving up the product stack is only an option if you have no specific budget (price point) in mind.
Disappointment is just the result of a collision between unrealistic expectations and reality.Atm I'm slightly disappointed with how the new Ryzens are looking...
8C16T for $200 isn't unrealistic, AMD could easily have done it, but it decided to get Intel-style greedy by raising its gross profit target to 45%.Disappointment is just the result of a collision between unrealistic expectations and reality.